In some cases the %variance values for the 3D PLS-DA plot is not in descending order of the orthogonal components. For example component 3’s % is larger than component 2 (see example). Shouldn’t the %var explained decrease for each additional component?
